There is an urgent need for developing new targeted therapies for tumors where none currently exist, and for developing companion diagnostics to identify which patients will respond to the targeted therapy. Targeted therapeutics are making a significant impact in cancer treatment, but many patients with ovarian, triple- negative breast, pancreatic, or lung cancer do not yet have a reliable targeted therapeutic option. In clinical samples from each of these four cancer types, mesothelin (MSLN) is a tumor biomarker frequently expressed at high levels on the surface of tumor cells and correlating with poor prognosis. MSLN has limited expression in healthy tissues. Consequently, MSLN has broad potential as a novel tumor target for diagnosis and therapy. Thus, there is a critical need for MSLN-targeted therapeutics and molecular diagnostics to identify patients who are most likely to respond to MSLN-targeting therapies, although currently no such FDA-approved agents exist. We propose to engineer highly stable mesothelin targeting agents using the fibronectin non-antibody protein scaffold, for use as both a targeted therapeutic and molecular diagnostic. Targeting agents that can serve both diagnostic and therapeutic, or theranostic, roles have particular clinical promise, allowing diagnostic identification of patients most likely to respond o the partner therapeutic. The strategy of using theranostics has not yet entered routine clinical use, largely due to the lack of appropriate targeting molecules. To develop candidate theranostics, we will engineer a highly stable protein scaffold to bind MSLN with high specificity. The protein scaffold we will use is based on a domain of human fibronectin, whose biophysical properties make it attractive for both diagnostic and therapeutic applications. The interaction of MSLN with tumor cell surface biomarker MUC16 leads to increased tumor invasiveness and metastasis, so that blocking this interaction can have a direct therapeutic effect. We will engineer proteins specifically to block the binding interaction of MSLN and MUC16. In preliminary studies, we have used directed evolution and yeast-surface display to identify a pool of protein variants that bind with moderate affinity to the domain of MSLN that mediates its binding to MUC16. In Aim 1, we will introduce further genetic diversity into our library of proteins, and use directed evolution to engineer proteins that bind to MSLN with high affinity and block its activation by MUC16. In Aim 2, we will measure clinically relevant biophysical properties of engineered proteins, including binding to tumor cells expressing MSLN. In Aim 3, we will measure in vitro bioactivity of engineered proteins, assessing their ability to inhibit cel migration and activate cell death. We will test the engineered proteins as in vivo molecular imaging agents, measuring biodistribution and tumor contrast in mouse xenograft models. Upon completion of this proposal, we will have identified therapeutic/diagnostic proteins targeting MSLN. A broad outcome of our proposed work is to validate the fibronectin protein scaffold for development of other targeted therapeutic/diagnostic molecules.